When we look at the progression of AI, specifically in the Command and Conquer franchise, we can see that across very few years a lot has happened to make the computer-based opponents more intelligent and more challenging, to better equip them with faster, strategic unit production, tactical attack waves and base building. All such that for the player, they feel a more realistic challenge and that these computer opponents feel more human and they feel more challenging. Balance is still important here: while the AI must become better, the player cannot have the feeling that the AI is cheating or has more and better resources than the player themselves; this would break immersion and cause frustration.
Command and Conquer has a legacy, and we want to focus on three distinct titles when we look at the evolution of the C&C AI: Command and Conquer 1, which is Tiberian Dawn, Command and Conquer 2 Red Alert, and the predecessor of both of them, which is Dune 2, the first real-time strategy game from 1993. Many of the mechanisms, mechanics, UI and gameplay that we know and love in Command and Conquer were already established or at least built in sometimes slightly more foundational logic, sometimes less complex, in Dune 2. When in Dune 2, we engage with opposing other fractions called “Houses”, often violently, to gain control over the planet, to gain control over the valuable “spice”, that in turn, seems to control the galaxy. When you engage with opposing houses, you will find that these Houses have some very rudimentary intelligence that for Dune 2 is mostly based on simple logic that is driven by scripts. These scripts determine – per level and scenario – when the computer intelligence should start building a specific thing, harvest a specific thing, build new units, and schedule its attack waves. There’s no sophisticated logic that specifically tells units to scout something, and there is no conclusions drawn from the map or the worldview, what needs to be done next, to challenge the player – most actions are scripted.
That changed a little bit. with Command and Conquer 1. Between the Dune 2 and C&C, there is two years of development. We see that Westwood Studios changed some of the AI logic away from purely scripts to some logic that inspects the game world and inspects the current state of units, buildings, and resources, and then draws conclusion and next actions from them. So there’s a clear shift in C&C going away from script-based and script-only actions that the computer opponent is carrying out, to inspecting the game world, if only for a few data points, to drawing conclusions based on them.
For example, there is functions that inspect if the computer opponent has sufficient resources to build additional units. And there’s a succession of units that need to be built. There’s also logic that determines if the computer opponent should give up, start its last offensive attack wave before it waves the white flag and will consider itself defeated by selling off all of the remaining buildings and then leveraging the units that spawned from the sold buildings to attack the human opponents.
There’s also logic to do more clever path finding such that units no longer run themselves into situations where they can’t get out and things like that. The world is inspected such that the AI knows how many units it currently has and how many resources it has and makes then decisions on what it should build next and how it should react to events on the battlefield, i.e. it will more cleverly fight back or run away from a specific situation or schedule attack waves based on the units of the resources that it already has. So it’s less about time schedules, more about it has harvested enough resources, it has built enough units, now is the time to build a small group and then go past the user. the player.
For Red Alert, we see yet another evolution of the game AI more and more towards exploration of the game world and even more introspection about what the AI should be doing next. We’re considering more data points and more actions than we did in Tiberian Dawn.
Red Alert for example, adds two additional properties into its AI system that determine how the AI is behaving: one is called TechLevel the other is called the house IQ. The TechLevel is new in Red Alert and didn’t exist in Tiberian Dawn yet. The idea of TechLevel is to track which technologies, units and buildings the AI can create and built in any given scenario or mission. It is a way of defining how far the tech tree is open to the AI system and how much the AI can build in any given scenario for the single player scenarios. The computer-based opponent would get its TechLevel assigned through the mission or through the scenario scripting and would have to stick to that tech level throughout that scenario. The higher the TechLevel, the more advanced units and buildings and weapons the AI can build over time with its resources. So the ability to build cleverly is tied to the TechLevel that the developers put in the mission script into Red Alert and every mission can have a different starting position and therefore openness of the tech tree for the AI opponent. Like this, different scenarios and AI cleverness, as well as pressure on the player can be created – but varying the weapons and units the AI has at its disposal.
The other property in the AI system that Red Alert adds, is the house IQ. The idea of the IQ, is defining which abilities the AI opponent has in any given scenario similar to the TechLevel or the BuildLevel. The IQ defines what options the AI can have and execute in any given scenario, such as how fast and how elaborate the AI can build and produce units, how the AI plans their attack waves, how aggressive or how clever scouting of the map throughout the scenario happens. It also steers how the AI reacts when it is under attack and how cleverly it guards specific buildings or specific units challenging the player even further.
So with these additional levels and ways to fine-tune the AI mission by mission and scenario by scenario gives the designers far greater flexibility in fine-tuning how the AI behaves and therefore make it more believable that the AI is an opponent and doesn’t cheat – and it makes it a little less predictable than the previous games.
FRICKELSOFT 